whereismyai
mini-SWE-agent
mini-SWE-agent is an AI coding agent that solves GitHub issues with 100 lines of code while maintaining high performance on the SWE-bench benchmark.
mini-SWE-agent is a minimal AI software engineering agent designed to solve GitHub issues. Built by the research team behind SWE-bench and SWE-agent, it uses approximately 200 lines of code in total across environment, model, and script components to reduce complexity. The agent supports all language models through litellm, openrouter, and portkey integrations, including features like interleaved thinking and various API endpoints. It runs in local environments or via containerization tools like docker and podman. The agent is intended for researchers and developers seeking a simplified, efficient alternative to larger agentic frameworks.
mini-SWE-agent at a glance
- Best for: Software engineer
- Runs on: CLI, self-hosted
- Pricing: Free
- Released: June 2025
- Category: AI coding tool
- Website: mini-swe-agent.com
Key features
- Scores over 74% on the SWE-bench verified benchmark
- Uses approximately 100 lines of code for the agent
- Supports all models via litellm, openrouter, and portkey
- Includes Python bindings for custom development
- Supports local environments, docker, podman, and singularity
- Offers batch inference capabilities
- Provides a trajectory browser
Use it for
- Solving GitHub issues
- Running batch inference tasks
- Building minimal AI agents
- Conducting research on agent performance
Companions
Tools that articles about mini-SWE-agent recommend using with it, each with its job.
- Context7 — Provides up-to-date documentation for AI agents and code editors.
- Ollama — Run local models at zero marginal cost, with code kept on your machine.
- OpenRouter — Access over 500 models and 80 providers through one unified API.
- Daytona — Runs sessions in isolated sandboxes synced to git branches.
- Oh My OpenCode Slim — Lightweight multi-agent orchestration and background tasks.
Alternatives to mini-SWE-agent
- SmallCode — Runs fully local with budget-managed context and a forgiving parser for small models.
- Open SWE — It uses isolated cloud sandboxes and curated toolsets for production engineering.
- Little Coder — Optimized for small local models with built-in planning and research tools.
- jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
- OpenCode — Works with 75+ model providers, local models, and existing subscriptions
- GoCodeo — Generates production-ready code, explains code, and creates tests in seconds.
- AG-UI — Standardizes agent-user interactions and generative UI for applications.
- Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
- Git AI — It moves work to the server and handles concurrent tasks without standard Git bottlenecks.
- Mcode Stack — Provides real everyday engineering habits and skills for AI coding agents.
- Lody — Run coding agents in parallel with isolated worktrees and live diff reviews.
- TRAE — It features an autonomous coding agent that drives the entire development process.
All mini-SWE-agent alternatives